A control system and method for a production line of a gypsum-based fire-retardant board based on a PLC

Abstract

The application relates to the technical field of automatic production control and industrial data processing, in particular to a gypsum-based fire-retardant plate production line control system and method based on PLC, which comprises the following modules: a data acquisition module for acquiring sensor data in real time; an ideal state construction module for constructing a pure ideal state vector; a disturbance simulation module for generating a theoretical damaged state vector corresponding to a specific fault mode; a difference calculation module for generating a real and theoretical deviation vector; a coupling verification control module for calculating the similarity value between the real deviation vector and the theoretical deviation vector; if the similarity value is greater than a preset judgment threshold, it is determined that a real physical fault exists at present, and a PLC precise compensation instruction is generated and sent to a production line actuator; otherwise, it is determined that a non-physical noise exists at present, the production line control parameters remain unchanged, and a dynamic filtering program is started; the application effectively solves the problem that physical faults and sensor clutter are difficult to distinguish in a high-noise environment, and avoids false negatives.

Description

Technical Field

[0001] This invention relates to the field of automated production control and industrial data processing technology, specifically to a PLC-based control system and method for a gypsum-based flame-retardant board production line. Background Technology

[0002] In the current automated production process of gypsum-based flame-retardant boards, the PLC control system needs to periodically process a large amount of real-time sensor data from the mixing system, molding station and drying kiln, etc. These data are affected by multiple factors such as fluctuations in raw material purity, changes in ambient temperature and humidity and random noise caused by the electromagnetic environment of the production line, exhibiting high nonlinearity and uncertainty.

[0003] To maintain stable production, existing control schemes generally adopt logic based on out-of-tolerance alarms, which monitor parameters such as temperature, pressure, or current by setting fixed thresholds. Although this scheme has a certain monitoring capability when dealing with obvious equipment failures, it lacks a deep understanding of the physical mechanisms and cannot effectively distinguish between random noise from sensors and weak early physical faults. This limitation makes the system prone to false alarms in high-noise environments, or it may miss the best maintenance opportunity due to the inability to identify physical anomalies in their early stages, resulting in frequent and invalid fluctuations in control parameters, making it difficult to support the high-precision adaptive production requirements. Therefore, how to accurately identify weak physical faults in strong noise environments and eliminate non-physical interference caused by environmental fluctuations, thereby improving the accuracy and operational stability of adaptive control of the production line, has become an urgent technical problem to be solved. Summary of the Invention

[0004] The purpose of this invention is to provide a PLC-based control system and method for a gypsum-based flame-retardant board production line. This aims to solve the problems of difficulty in identifying subtle physical faults under strong noise conditions and the susceptibility of false alarms caused by environmental fluctuations in the control system. Through a digital twin model driven by physical mechanisms and dual-track differential comparison technology, it achieves precise decoupling between real physical faults and non-physical noise, thereby improving the accuracy and operational stability of the production line's adaptive control. Specifically, the technical solution of this invention is as follows:

[0005] A PLC-based control system for a gypsum-based flame-retardant board production line includes:

[0006] The data acquisition module is configured to collect data on the purity of raw materials, environmental temperature and humidity parameters, and real-time sensor data of the PLC at each station of the gypsum-based flame-retardant board production line in real time through the analog input interface of the PLC.

[0007] The ideal state construction module is configured to substitute the raw material purity data and the environmental temperature and humidity parameters into the preset gypsum hydration reaction kinetic equation, and deduce and calculate the theoretical numerical sequence of each sensor changing over time under fault-free conditions, so as to construct a pure ideal state vector.

[0008] The disturbance simulation module is configured to call a preset fault mechanism model from the fault parameter database, inject the quantified fault factor parameters into the gypsum hydration reaction kinetic equation, and generate a theoretical damaged state vector corresponding to a specific fault mode.

[0009] The differential calculation module is configured to perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector, and to perform vector subtraction on the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector;

[0010] The coupling verification control module is configured to calculate the similarity value between the actual deviation vector and the theoretical deviation vector; if the similarity value is greater than a preset judgment threshold, it is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator; if the similarity value is less than or equal to the preset judgment threshold, it is determined that there is non-physical noise, the production line control parameters are kept unchanged, and a dynamic filtering program is started.

[0011] Preferably, the ideal state construction module performs the construction of the pure ideal state vector, specifically including:

[0012] The sensor reads the current percentage of hemihydrate gypsum content in the raw material, the ambient temperature value, the ambient humidity value, and the water-to-gypsum ratio setting value of the production line.

[0013] The read data is used as boundary conditions and substituted into the preset gypsum hydration reaction kinetic equation and thermodynamic heat transfer equation.

[0014] By numerical integration, the ideal operating state of the production line within a set time window is derived, generating a pure ideal state vector containing the dimensions of temperature, pressure, and flow.

[0015] Preferably, the perturbation simulation module performs the generation of the theoretical damaged state vector, specifically including:

[0016] Extract the mechanism model of a specific fault type from the preset fault parameter database, and transform the mechanism model into a quantization perturbation operator that acts on the equation coefficients;

[0017] The perturbation operator is superimposed onto the reaction rate constant or heat transfer coefficient of the gypsum hydration reaction kinetic equation, and lossy simulation calculation is performed.

[0018] Output the theoretical damaged state vector of each sensor value over time under this specific fault type.

[0019] Preferably, the difference calculation module performs the following operations:

[0020] The PLC real-time sensor data and the pure ideal state vector are subjected to element-wise vector subtraction to obtain the actual deviation vector containing noise components and potential fault signal components.

[0021] By performing element-wise vector subtraction between the theoretical damaged state vector and the pure ideal state vector, a theoretical deviation vector characterizing a specific fault numerical feature pattern is obtained.

[0022] Preferably, the calculation of similarity values ​​by the coupling verification control module specifically includes:

[0023] The actual deviation vector and the theoretical deviation vector are mapped to a multidimensional Euclidean space composed of sensor data dimensions;

[0024] Calculate the cosine of the angle between the actual deviation vector and the theoretical deviation vector, and use the cosine as the similarity value; read the judgment threshold preset in the PLC memory, which is a value determined based on the statistical distribution characteristics of historical fault data;

[0025] When the cosine value is strictly greater than the judgment threshold, a real physical fault confirmation signal is generated to trigger the PLC's precise compensation operation.

[0026] When the cosine value is less than or equal to the determination threshold, a non-physical noise confirmation signal is generated to trigger dynamic filtering operation.

[0027] Preferably, the failure mechanism models include the nozzle clogging model, the agitator shaft wear model, and the flame retardant uneven distribution model;

[0028] The nozzle clogging model is defined as a step decay function of the flow coefficient over time, characterizing the abrupt change in fluid resistance; the agitator shaft wear model is defined as a specific frequency harmonic function superimposed on the motor torque output signal, characterizing the periodic vibration caused by mechanical eccentricity.

[0029] The uneven distribution model of the flame retardant is defined as a discretized variance-increasing function of the thermal conductivity parameter of the local area of ​​the board, which characterizes the random discreteness of the material distribution.

[0030] Preferably, the coupling verification control module performs PLC precise compensation operations, including:

[0031] Identify the fault type corresponding to the theoretical damaged state vector, and call the PID parameter correction table pre-bound to the fault type from the strategy library;

[0032] Based on the magnitude of the actual deviation vector, a reverse control command is generated to adjust the operating parameters of the production line actuator to counteract the deviation caused by the actual physical fault.

[0033] Preferably, the coupling verification control module performs dynamic filtering operations, including:

[0034] The actual deviation vector is identified as random noise, and the control parameters of the PLC are not adjusted.

[0035] The sensor data collected subsequently are smoothed using a moving average filtering algorithm or a Kalman filtering algorithm to eliminate data fluctuations.

[0036] A PLC-based control method for a gypsum-based flame-retardant board production line includes the following steps:

[0037] Data acquisition steps: Real-time data acquisition of raw material purity, environmental temperature and humidity parameters, and PLC real-time sensor data at each station of the gypsum-based flame-retardant board production line is conducted through the PLC's analog input interface.

[0038] Ideal state construction steps: The raw material purity data and the environmental temperature and humidity parameters are used as dynamic boundary conditions. They are substituted into the gypsum hydration reaction kinetic equation and related physical field equation preset in the PLC storage area to deduce and calculate the theoretical numerical sequence of each sensor changing with time under fault-free conditions, so as to construct a pure ideal state vector.

[0039] Disturbance simulation steps: Call the preset fault mechanism model from the fault parameter database, inject the quantified fault factor parameters into the gypsum hydration reaction kinetic equation and the associated physical field equation, perform loss-bearing operation simulation in the virtual environment, and generate the theoretical damaged state vector corresponding to the specific fault mode.

[0040] Differential calculation steps: Perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector, and perform vector subtraction on the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector;

[0041] Coupling verification control steps: Calculate the similarity value between the actual deviation vector and the theoretical deviation vector, and execute the following judgment logic:

[0042] If the similarity value is greater than the preset judgment threshold, it is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator.

[0043] If the similarity value is less than or equal to the preset judgment threshold, it is determined that there is non-physical noise, the production line control parameters remain unchanged, and the dynamic filtering program is started.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. This invention achieves accurate identification of early and weak faults under strong noise background; it changes the traditional out-of-tolerance alarm logic by constructing a pure ideal state vector as a dynamic reference system and using the deviation template generated by the fault mechanism model for waveform similarity matching; this method can accurately capture weak early signals such as slight nozzle blockage or initial wear of the stirring shaft, effectively solving the problem of difficulty in distinguishing physical faults from sensor noise in high noise environment and avoiding missed alarms.

[0046] 2. This invention eliminates environmental fluctuation interference and significantly reduces the false alarm rate of the system; the system substitutes dynamic boundary conditions such as raw material purity and ambient temperature and humidity into the first-principles physical equations to deduce the physically achievable ideal state in real time; even under extreme conditions such as cold waves or raw material batch fluctuations, the system can automatically update the benchmark and judge the parameter deviation caused by the environment as normal physical evolution rather than a fault, thereby completely shielding non-physical noise and reducing invalid alarms and false shutdowns;

[0047] 3. This invention has the ability to identify and warn of rare faults with zero samples; the disturbance simulation module can reverse-engineer and generate fingerprint features of specific faults by injecting quantified fault factors into the dynamics and heat transfer equations; this means that even if a certain fault has never occurred on the production line, as long as its physical mechanism is known, the system can predict its data performance and match it through simulation; this greatly improves the detection rate of rare but fatal faults and provides a scientific basis for preventive maintenance.

[0048] 4. This invention improves the adaptive control accuracy of production lines operating with defects; after confirming a physical fault, the coupled verification control module can identify the fault type and call the corresponding parameter correction table, and generate accurate compensation instructions through the damped least squares method; this strategy of combining feedforward and feedback can offset the negative impact of physical faults by adjusting the actuator parameters without immediately stopping the machine, maintain product quality stability, and ensure production continuity. Attached Figure Description

[0049] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0050] Figure 1 This is a structural diagram of the system of the present invention;

[0051] Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0053] Example 1:

[0054] Please see Figure 1 A PLC-based control system for a gypsum-based flame-retardant board production line includes:

[0055] The data acquisition module is configured to collect data on the purity of raw materials, environmental temperature and humidity parameters, and real-time sensor data of the PLC at each station of the gypsum-based flame-retardant board production line in real time through the analog input interface of the PLC.

[0056] The ideal state construction module is configured to substitute the raw material purity data and the environmental temperature and humidity parameters into the preset gypsum hydration reaction kinetic equation and related physical field equation, and deduce and calculate the theoretical numerical sequence of each sensor changing with time under fault-free conditions, so as to construct a pure ideal state vector.

[0057] The disturbance simulation module is configured to call a preset fault mechanism model from the fault parameter database, inject the quantified fault factor parameters into the gypsum hydration reaction kinetic equation and the associated physical field equation, and generate a theoretical damaged state vector corresponding to a specific fault mode.

[0058] The differential calculation module is configured to perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector, and to perform vector subtraction on the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector;

[0059] The coupling verification control module is configured to calculate the similarity value between the actual deviation vector and the theoretical deviation vector; if the similarity value is greater than a preset judgment threshold, it is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator; if the similarity value is less than or equal to the preset judgment threshold, it is determined that there is non-physical noise, the production line control parameters are kept unchanged, and a dynamic filtering program is started.

[0060] This embodiment provides a PLC-based control system for a gypsum-based flame-retardant board production line. The system is based on analysis-synthesis closed-loop control logic, which aims to solve the technical problem that traditional control systems have difficulty distinguishing between sensor noise and weak physical faults.

[0061] Data acquisition module: This is defined as the interface layer between the system and the physical world; in this embodiment, the module is configured to acquire the following three types of data in real time at millisecond intervals via the PLC's high-speed analog input interface:

[0062] Raw material purity data: refers to the hemihydrate gypsum entering the mixing system. The percentage of effective content, which is usually determined by an online near-infrared spectrometer;

[0063] Ambient temperature and humidity parameters: These refer to the air temperature and relative humidity around the production line. These parameters directly affect the initial setting time of the gypsum hydration reaction.

[0064] PLC real-time sensor data refers to the real-time readings of temperature, pressure, flow rate, and motor current sensors installed on key workstations in the production line, such as mixers, molding stations, and drying kilns.

[0065] Ideal State Construction Module: Defined as a first-principles-based digital twin reference generator; unlike traditional references based on historical statistical data, this module is configured to use the raw material purity data and the environmental temperature and humidity parameters as dynamic boundary conditions, substituting them into the gypsum hydration reaction kinetic equations and related physical field equations preset in the PLC storage area; through numerical solution, this module derives and calculates the theoretical numerical sequence of each sensor change over time under the assumption of completely fault-free equipment and perfect process; after standardization, this sequence is constructed into a pure ideal state vector. ;

[0066] The disturbance simulation module is defined as a reverse inference engine based on fault mechanisms. This module is configured to call preset fault mechanism models, such as blockage and wear mathematical descriptions, from a fault parameter database. The system injects quantified fault factor parameters, such as specific attenuation of the flow resistance coefficient, into the aforementioned gypsum hydration reaction kinetic equations and related physical field equations. It then performs loss-carrying operation simulations in a virtual environment, generating theoretical damaged state vectors corresponding to specific fault modes. ;

[0067] Difference calculation module: It is defined as a feature extraction operator; this module is configured to perform dual-track difference operations.

[0068] First track: Perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector. This vector mixes real physical fault signals with non-physical random noise.

[0069] Second track: Perform vector subtraction between the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector. This vector purely represents the fingerprint characteristics of a specific fault type and contains no environmental noise.

[0070] Coupled verification control module: It is defined as the core of the system's decision-making and execution; this module is configured to calculate the actual deviation vector. The theoretical deviation vector The similarity value between them;

[0071] Judgment logic: If the similarity value is greater than the preset judgment threshold, it indicates that the actual deviation pattern matches the main characteristics of the theoretically deduced fault pattern. It is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator, such as adjusting the frequency converter frequency or valve opening. If the similarity value is less than or equal to the preset judgment threshold, it indicates that the actual deviation is caused by random interference. It is determined that there is non-physical noise, and the production line control parameters are kept unchanged and the dynamic filtering program is started.

[0072] By introducing a pure ideal state vector as a dynamic coordinate system and using a theoretical damaged state vector as a comparison template, this system achieves a leap from numerical out-of-tolerance alarms to physical semantic understanding. Its core advantage lies in that it can not only identify significant and obvious faults, but also accurately identify weak early physical faults, such as slight nozzle blockage, through waveform similarity matching in a strong noise background. At the same time, it completely shields false alarms caused by environmental noise, significantly improving the adaptive control accuracy and operational stability of the production line.

[0073] Example 2:

[0074] The ideal state construction module performs the construction of the pure ideal state vector, specifically including:

[0075] The sensor reads the current percentage of hemihydrate gypsum content in the raw material, the ambient temperature value, the ambient humidity value, and the water-to-gypsum ratio setting value of the production line.

[0076] The read data is used as boundary conditions and substituted into the preset gypsum hydration reaction kinetic equation and thermodynamic heat transfer equation.

[0077] By numerical integration, the ideal operating state of the production line within a set time window is derived, generating a pure ideal state vector containing the dimensions of temperature, pressure, and flow.

[0078] This embodiment is a further specification of the ideal state construction module in Embodiment 1, focusing on solving the black box problem of the gypsum hydration reaction kinetic equation and thermodynamic heat transfer equation, so as to meet the requirement of sufficient disclosure;

[0079] In this embodiment, the construction process of the pure ideal state vector strictly follows the following well-defined mathematical and physical model:

[0080] The sensor reads the current purity of hemihydrate gypsum. The unit is %, ambient temperature. The unit is K, and the ambient humidity is... The units are % and the water-to-solid ratio set value. ;

[0081] The system uses the following specific set of differential equations to describe the reaction process: hydration kinetic equations, the calculation formula of which is:

[0082]

[0083] in, The degree of hydration is 0~1; This is the diffusion resistance coefficient; The reaction rate constant is defined by the Arrhenius equation:

[0084]

[0085] The input parameters are specified here. The correction coefficient, used as a pre-exponential factor, participates in the calculation, where, The pre-calibrated reference exponential factor has dimensions that are different from the reaction rate constant. Consistent, and Substitute 95% into the calculation as a decimal, such as 0.95; For activation energy, Let be the ideal gas constant, and take the value of . ;

[0086] The thermodynamic heat transfer equation, its calculation formula is:

[0087]

[0088] Here The effective volume of the reaction zone. For slurry density, is the specific heat capacity; the left side of the equation represents the rate of change of the total sensible heat of the system; the first term on the right side is the exothermic power of the chemical reaction. The first term represents the heat of reaction per unit mass, and the second term represents the convective heat transfer power. The convective heat transfer coefficient is... For effective heat exchange area, the third item To be affected by ambient humidity The impact of latent heat loss on power; based on this, in order to achieve code-level reproduction, the latent heat loss power... The calculation logic is as follows: The saturated vapor pressure at the current temperature is calculated using the Antoine equation. ,in, , , The pre-defined Antoine constant for water vapor; based on Dalton's law of evaporation, the latent heat loss is quantified as:

[0089] in, The mass transfer coefficient is experimentally measured, and its value range is usually [value range missing]. to The specific values ​​are obtained by calibration based on the exhaust air velocity and spatial geometry of the production line. Let be the latent heat of vaporization of water;

[0090] Before performing numerical integration, set initial boundary conditions: Let Hydration at time And the initial theoretical temperature Equal to the current ambient temperature read by the sensor The above system of differential equations was numerically integrated using the fourth-order Runge-Kutta method (RK4), with a time step of [missing information]. Points earned Theoretical temperature at time Compared with theoretical hydration degree Then, the theoretical viscosity is calculated using the equation of state. Theoretical stirring torque With theoretical pressure :

[0091]

[0092]

[0093]

[0094] in, This is the equivalent length of the slurry delivery pipeline in the production line. Pipe inner diameter; pump characteristic function Defined as based on centrifugal pump The formula for calculating the quadratic fitting polynomial of the curve is:

[0095] in, These are constant coefficients that were calibrated in advance using pump pressure test data; For the geometric factors of the mixer, This refers to the stirring speed. This is the pump characteristic function;

[0096] Finally, a pure ideal state vector containing multidimensional physical quantities is constructed. To ensure dimensional matching with the sensor acquisition module containing motor current data, it should be noted that, in order to eliminate the dimensional difference between the motor current acquired by the sensor and the theoretically calculated torque, the system performs a consistency transformation before constructing the vector: using the motor torque constant. Convert theoretical torque into theoretical current Correspondingly, the pure ideal state vector participating in the subtraction operation is corrected to... This allows for a one-to-one correspondence between the dimensions of real-time sensor data;

[0097] By constructing a benchmark through real-time solution of physical equations, this embodiment abandons the static standard curve used in traditional methods. This means that even in extreme weather, such as when a cold wave causes a sudden drop in ambient temperature or when raw material batches fluctuate greatly, the ideal state generated by the system is still the best physically achievable state, thereby avoiding control deviations caused by the benchmark itself being unable to adapt to environmental changes.

[0098] Example 3:

[0099] The perturbation simulation module performs the generation of the theoretical damaged state vector, specifically including:

[0100] Extract the mechanism model of a specific fault type from the preset fault parameter database, and transform the mechanism model into a quantization perturbation operator that acts on the equation coefficients;

[0101] The perturbation operator is superimposed onto the coefficients or state variables of the gypsum hydration reaction kinetic equation and the associated physical field equation, and lossy simulation calculation is performed.

[0102] Output the theoretical damaged state vector of each sensor value over time under this specific fault type.

[0103] This embodiment supplements the specific mathematical expression of the flame retardant uneven distribution model, ensuring that all fault models are programmable. The flow resistance step of the nozzle blockage model is calculated using the following formula:

[0104]

[0105] in, The initial flow coefficient, To be at the moment of failure The unit step function, The preset blockage coefficient is 0.1~0.5;

[0106] The calculation formula for the torque harmonic superposition of the agitator shaft wear model is as follows:

[0107]

[0108] in, As the reference torque, This refers to the real-time operating frequency of the motor. This is the eccentric harmonic coefficient, which can be either 1 or 2. This represents the wear amplitude. This is the initial phase;

[0109] Discretized variance injection for flame retardant uneven distribution model: This model is defined as a spatial random perturbation of the thermal conductivity parameter of the sheet material; in the simulation calculation, the discretized thermal conductivity... Revised to:

[0110]

[0111] in, The nominal thermal conductivity, It is a random noise sequence that follows a Gaussian distribution, and its calculation formula is:

[0112]

[0113] This is the uneven distribution factor; when simulating this fault, the system will... Increasing the value from the default value of 0.01 to 0.15 and substituting it into the thermodynamic equation results in a non-smooth high-frequency oscillation characteristic of the temperature component in the theoretical damaged state vector, thereby characterizing the random discreteness of the material distribution.

[0114] By introducing a fault model based on physical mechanisms, this system has zero-sample identification capability. Even if a certain fault has never occurred on the production line, as long as its physical mechanism, such as vibration caused by wear, is known, the system can predict its manifestation in sensor data through simulation, thereby greatly improving the detection rate of rare but fatal faults.

[0115] Example 4:

[0116] The difference calculation module performs the following operations:

[0117] The PLC real-time sensor data and the pure ideal state vector are subjected to element-wise vector subtraction to obtain the actual deviation vector containing noise components and potential fault signal components.

[0118] By performing element-wise vector subtraction between the theoretical damaged state vector and the pure ideal state vector, a theoretical deviation vector characterizing a specific fault numerical feature pattern is obtained.

[0119] This embodiment is a further specification of the difference calculation module;

[0120] In this embodiment, differential computation serves as a bridge connecting the physical world and the digital world, and the specific operation is as follows:

[0121] Generate reality deviation vector The calculation formula is as follows:

[0122]

[0123] This subtraction operation is performed element-by-element; It contains real-time sensor data from the PLC, which includes noise.

[0124] Results analysis: It not only contains potential fault signals, but also inevitably contains measurement noise and environmental interference;

[0125] Generate theoretical deviation vector The calculation formula is as follows:

[0126]

[0127] This operation is also performed element by element;

[0128] Results analysis: It is a pure mathematical object that represents the pure shape that a specific fault should theoretically exhibit, that is, the numerical characteristic pattern of a specific fault, completely free of random noise;

[0129] By using double difference, the system transforms the problem from absolute value comparison to deviation pattern comparison; this approach eliminates the influence of the static operating point, allowing the system to focus only on the variable. As a noise-free reference template, it provides a high signal-to-noise ratio benchmark for extracting weak fault features in a high-noise environment.

[0130] Example 5:

[0131] The calculation of similarity values ​​by the coupling verification control module specifically includes:

[0132] The actual deviation vector and the theoretical deviation vector are mapped to a multidimensional Euclidean space composed of sensor data dimensions;

[0133] Calculate the cosine of the angle between the actual deviation vector and the theoretical deviation vector, and use the cosine value as the similarity value;

[0134] Read the judgment threshold preset in the PLC memory. The judgment threshold is a value determined based on the statistical distribution characteristics of historical fault data.

[0135] When the cosine value is strictly greater than the judgment threshold, a real physical fault confirmation signal is generated to trigger the PLC's precise compensation operation.

[0136] When the cosine value is less than or equal to the determination threshold, a non-physical noise confirmation signal is generated to trigger dynamic filtering operation.

[0137] This embodiment is a further specification of the construction of similarity values ​​and verification logic in the coupled verification control module;

[0138] In this embodiment, in order to accurately determine Does it contain A high-dimensional geometric projection method is used:

[0139] Spatial mapping: mapping the actual deviation vector The theoretical deviation vector Mapped to dimensions of sensor data, such as temperature axis, pressure axis, flow axis, torque axis, etc. Vioclimatic space;

[0140] Cosine similarity calculation: To eliminate the influence of signal amplitude, such as the severity of the fault, and focus on waveform trend matching, this embodiment uses the cosine value of the included angle as the similarity value. The calculation formula is as follows:

[0141]

[0142] in, It is a preset minimal constant used to avoid calculation anomalies when the denominator is zero;

[0143] The dot product of two vectors;

[0144] and The L2 norm of two vectors;

[0145] The value ranges from [-1, 1]; the closer it is to 1, the more consistent the two are in their orientation in multidimensional space, meaning that the evolution law of the actual deviation is highly consistent with the evolution law of the theoretical fault; it should be noted that if Less than the preset minimum constant ,like Then the system will make a mandatory judgment. This is to avoid the anomaly of a zero denominator caused by excessively small deviations from reality;

[0146] Threshold determination and signal generation:

[0147] The system reads the preset judgment threshold from the PLC memory. This threshold is determined based on the statistical distribution characteristics of historical fault data. For example, the 5th percentile of the similarity of confirmed fault samples in history is taken, usually set to 0.85-0.95.

[0148] when Time: System confirmation The fluctuations in the data are caused by physical faults, generating a real physical fault confirmation signal;

[0149] when Time: System confirmation The fluctuations in the signal are random clutter or unstructured interference, generating non-physical noise confirmation signals;

[0150] Using cosine similarity instead of Euclidean distance as the core criterion makes the system sensitive to the shape of the fault but relatively insensitive to the magnitude of the fault. This means that even if the fault is in its infancy, as long as the trend of its multidimensional data changes conforms to physical laws, the system can identify it. At the same time, for those sensor jumps with large amplitudes but that do not conform to physical laws, the system can resolutely determine them as noise, thereby effectively avoiding false shutdowns.

[0151] Example 7:

[0152] The coupling verification control module performs precise compensation operations on the PLC, including:

[0153] Identify the fault type corresponding to the theoretical damaged state vector, and call the PID parameter correction table pre-bound to the fault type from the strategy library;

[0154] Based on the magnitude of the actual deviation vector, a reverse control command is generated to adjust the operating parameters of the production line actuator to counteract the deviation caused by the actual physical fault.

[0155] This embodiment further specifies the precise compensation operation of PLC, clarifying the logical relationship and specific calculation formula between the PID parameter correction table and the reverse control command;

[0156] When a real physical fault is determined to exist, the system executes a control strategy that combines feedforward compensation and feedback parameter scheduling:

[0157] The system identifies the fault type, for example, ID=3: minor nozzle blockage, and reads the corresponding sensitivity Jacobian matrix from the PLC lookup table. This matrix describes control variables, such as pump speed. belt speed For sensor variables, such as pressure ,temperature The partial derivative relationship;

[0158] To offset the actual deviation vector Right now The system calculates the compensation control vector. The calculation formula is as follows:

[0159]

[0160] in, For matrix For identity matrices of the same dimension, the order is determined by the number of control variables; damped least squares method is used here to solve for pseudo-inverses to prevent matrix singularities. The regularization parameter is set to 0.01. The final instruction output to the actuator is:

[0161]

[0162] in, The output of the original PID loop is directly superimposed. The system does not change the PID parameters In such cases, a bias that can quickly offset the physical effects of the fault is introduced to maintain quality while the fault is in operation;

[0163] This embodiment achieves quality maintenance while operating with defects; for minor physical faults that cannot be immediately stopped, such as minor blockages or wear, the system can offset the impact of the fault on product quality within a certain range through multi-parameter collaborative compensation, ensuring the continuity of production until the planned maintenance window arrives.

[0164] Example 8:

[0165] The coupled verification control module performs dynamic filtering operations, including: identifying the actual deviation vector as random noise and not adjusting the control parameters of the PLC; and applying a moving average filtering algorithm or a Kalman filtering algorithm to smooth the subsequently acquired sensor data in order to eliminate data fluctuations.

[0166] This embodiment further specifies the dynamic filtering operation, clarifies how the similarity value specifically affects the parameters of the filtering algorithm, and establishes a closed-loop mathematical correlation.

[0167] When the system enters dynamic filtering mode, it employs adaptive Kalman filtering, the core of which lies in adjusting the measurement noise covariance matrix in real time based on similarity. ;

[0168] Adaptive law of covariance matrix: Let the covariance of the basic measurement noise be... The system calculates the cosine similarity based on the coupling verification module. ( Dynamically calculate the current measurement noise covariance. The calculation formula is as follows:

[0169]

[0170] in, : Preset judgment threshold; Gain coefficient, used to amplify noise weights; Sensitivity coefficient;

[0171] When similarity When the value is much smaller than the threshold, the exponential term increases sharply, leading to... It becomes extremely large; this means that the filter has extremely little trust in the current real-time measurements.

[0172] The formula for Kalman gain update and state estimation is as follows:

[0173]

[0174]

[0175] in, For the observation matrix, Let be the prior error covariance matrix. This represents the sensor measurement vector at the current moment. This is the state estimation vector; since The calculated Kalman gain is maximized. It will approach 0; at this point, the state update formula degenerates into... When performing the above recursive operation, the system's observation matrix Defined as identity matrix This represents a direct observation of physical parameters; simultaneously, to ensure the convergence of the filter, a pre-defined process noise covariance matrix is ​​used. A constant diagonal matrix Its element value range is to Between these, computational perturbations of the digital twin model itself are used to characterize the system's main data fluctuations; that is, the system mainly relies on the model's predicted values ​​from the previous moment, thereby effectively freezing the response to the current high-noise data and achieving smooth elimination of data fluctuations.

[0176] Unlike traditional global filtering, which can cause sluggish system response, this embodiment implements on-demand filtering; strong filtering logic is only activated when the system is certain that the current fluctuation has no physical meaning. This ensures both the system's rapid response to real changes during normal operation and the system's stability and robustness under noise interference.

[0177] Example 9:

[0178] Please see Figure 2 A PLC-based control method for a gypsum-based flame-retardant board production line includes the following steps:

[0179] Data acquisition steps: Real-time data acquisition of raw material purity, environmental temperature and humidity parameters, and PLC real-time sensor data at each station of the gypsum-based flame-retardant board production line is conducted through the PLC's analog input interface.

[0180] Ideal state construction steps: The raw material purity data and the environmental temperature and humidity parameters are used as dynamic boundary conditions. They are substituted into the gypsum hydration reaction kinetic equation and related physical field equation preset in the PLC storage area to deduce and calculate the theoretical numerical sequence of each sensor changing with time under fault-free conditions, so as to construct a pure ideal state vector.

[0181] Disturbance simulation steps: Call the preset fault mechanism model from the fault parameter database, inject the quantified fault factor parameters into the gypsum hydration reaction kinetic equation and the associated physical field equation, perform loss-bearing operation simulation in the virtual environment, and generate the theoretical damaged state vector corresponding to the specific fault mode.

[0182] Differential calculation steps: Perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector, and perform vector subtraction on the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector;

[0183] Coupling verification control steps: Calculate the similarity value between the actual deviation vector and the theoretical deviation vector, and execute the following judgment logic:

[0184] If the similarity value is greater than the preset judgment threshold, it is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator.

[0185] If the similarity value is less than or equal to the preset judgment threshold, it is determined that there is non-physical noise, the production line control parameters remain unchanged, and the dynamic filtering program is started.

[0186] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A PLC-based control system for a gypsum-based flame-retardant board production line, characterized in that, include: The data acquisition module is configured to collect data on the purity of raw materials, environmental temperature and humidity parameters, and real-time sensor data of the PLC at each station of the gypsum-based flame-retardant board production line in real time through the analog input interface of the PLC. The ideal state construction module is configured to substitute the raw material purity data and the environmental temperature and humidity parameters into the preset gypsum hydration reaction kinetic equation, and deduce and calculate the theoretical numerical sequence of each sensor changing over time under fault-free conditions, so as to construct a pure ideal state vector. The disturbance simulation module is configured to call a preset fault mechanism model from the fault parameter database, inject the quantified fault factor parameters into the gypsum hydration reaction kinetic equation, and generate a theoretical damaged state vector corresponding to a specific fault mode. The differential calculation module is configured to perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector, and to perform vector subtraction on the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector; The coupling verification control module is configured to calculate the similarity value between the actual deviation vector and the theoretical deviation vector; if the similarity value is greater than a preset judgment threshold, it is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator. If the similarity value is less than or equal to the preset judgment threshold, it is determined that there is non-physical noise, the production line control parameters remain unchanged, and the dynamic filtering program is started.

2. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 1, characterized in that, The ideal state construction module performs the construction of the pure ideal state vector, specifically including: The sensor reads the current percentage of hemihydrate gypsum content in the raw material, the ambient temperature value, the ambient humidity value, and the water-to-gypsum ratio setting value of the production line. The read data is used as boundary conditions and substituted into the preset gypsum hydration reaction kinetic equation and thermodynamic heat transfer equation. By numerical integration, the ideal operating state of the production line within a set time window is derived, generating a pure ideal state vector containing the dimensions of temperature, pressure, and flow.

3. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 1, characterized in that, The disturbance simulation module performs the generation of the theoretical damaged state vector, specifically including: Extract the mechanism model of a specific fault type from the preset fault parameter database, and transform the mechanism model into a quantization perturbation operator that acts on the equation coefficients; The perturbation operator is superimposed onto the reaction rate constant or heat transfer coefficient of the gypsum hydration reaction kinetic equation, and lossy simulation calculation is performed. Output the theoretical damaged state vector of each sensor value changing over time under this specific fault type.

4. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 1, characterized in that, The difference calculation module performs the following operations: The PLC real-time sensor data and the pure ideal state vector are subjected to element-wise vector subtraction to obtain the actual deviation vector containing noise components and potential fault signal components. By performing element-wise vector subtraction between the theoretical damaged state vector and the pure ideal state vector, a theoretical deviation vector characterizing a specific fault numerical feature pattern is obtained.

5. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 1, characterized in that, The calculation of similarity values ​​by the coupling verification control module specifically includes: The actual deviation vector and the theoretical deviation vector are mapped to a multidimensional Euclidean space composed of sensor data dimensions; Calculate the cosine of the angle between the actual deviation vector and the theoretical deviation vector, and use the cosine as the similarity value; read the judgment threshold preset in the PLC memory, which is a value determined based on the statistical distribution characteristics of historical fault data; When the cosine value is strictly greater than the judgment threshold, a real physical fault confirmation signal is generated to trigger the PLC's precise compensation operation. When the cosine value is less than or equal to the determination threshold, a non-physical noise confirmation signal is generated to trigger dynamic filtering operation.

6. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 3, characterized in that, The failure mechanism models include a nozzle clogging model, a stirring shaft wear model, and a flame retardant uneven distribution model. The nozzle clogging model is defined as a step decay function of the flow coefficient over time, characterizing the abrupt change in fluid resistance; the agitator shaft wear model is defined as a specific frequency harmonic function superimposed on the motor torque output signal, characterizing the periodic vibration caused by mechanical eccentricity. The uneven distribution model of the flame retardant is defined as a discretized variance-increasing function of the thermal conductivity parameter of the local area of ​​the board, which characterizes the random discreteness of the material distribution.

7. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 5, characterized in that, The coupling verification control module performs PLC precise compensation operations, including: Identify the fault type corresponding to the theoretical damaged state vector, and call the PID parameter correction table pre-bound to the fault type from the strategy library; Based on the magnitude of the actual deviation vector, a reverse control command is generated to adjust the operating parameters of the production line actuator to counteract the deviation caused by the actual physical fault.

8. The PLC-based control system for a gypsum-based flame-retardant board production line according to claim 5, characterized in that, The coupling verification control module performs dynamic filtering operations, including: The actual deviation vector is identified as random noise, and the control parameters of the PLC are not adjusted. The sensor data collected subsequently are smoothed using a moving average filtering algorithm or a Kalman filtering algorithm to eliminate data fluctuations.

9. A PLC-based control method for a gypsum-based flame-retardant board production line, characterized in that, Includes the following steps: Data acquisition steps: Real-time data acquisition of raw material purity, environmental temperature and humidity parameters, and PLC real-time sensor data at each station of the gypsum-based flame-retardant board production line is conducted through the PLC's analog input interface. Ideal state construction steps: The raw material purity data and the environmental temperature and humidity parameters are used as dynamic boundary conditions. They are substituted into the gypsum hydration reaction kinetic equation and related physical field equation preset in the PLC storage area to deduce and calculate the theoretical numerical sequence of each sensor changing with time under fault-free conditions, so as to construct a pure ideal state vector. Disturbance simulation steps: Call the preset fault mechanism model from the fault parameter database, inject the quantified fault factor parameters into the gypsum hydration reaction kinetic equation and the associated physical field equation, perform loss-bearing operation simulation in the virtual environment, and generate the theoretical damaged state vector corresponding to the specific fault mode. Differential calculation steps: Perform vector subtraction on the real-time sensor data of the PLC and the pure ideal state vector to generate the actual deviation vector, and perform vector subtraction on the theoretical damaged state vector and the pure ideal state vector to generate the theoretical deviation vector; Coupling verification control steps: Calculate the similarity value between the actual deviation vector and the theoretical deviation vector, and execute the following judgment logic: If the similarity value is greater than the preset judgment threshold, it is determined that there is a real physical fault, and a PLC precise compensation instruction is generated and sent to the production line actuator. If the similarity value is less than or equal to the preset judgment threshold, it is determined that there is non-physical noise, the production line control parameters remain unchanged, and the dynamic filtering program is started.

Images